Tuesday, October 20th 2009

Energy-Efficient Intel Core i5, Core i7 ''Lynnfield'' Processors Overclocked

In the run up for Intel's Core i5 and Core i7 "Lynnfield" socket LGA-1156 processors, it was known that Intel will be trailing the launch with a few energy-efficient variants, that come with lower rated TDPs. Two of these, namely Core i5 750S, and Core i7 860S have made it to Coolaler's lab for a quick check up. Unlike other energy-efficient variants from Intel, in the past, such as Core 2 Quad Q9550S, these parts feature lower clock-speeds compared to their standard counterparts.

The Core i5 750S comes with a clock speed of 2.40 GHz (19.0 x 133 MHz) with a Turbo Boost speed of 3.20 GHz, and the Core i7 860S runs at 2.53 GHz, with 3.46 GHz of Turbo Boost speed. Both processors as a result, have lower TDPs of 82W, compared to 95W on the Core i5 and Core i7 "Lynnfield" processors currently in the market. They are slated for Q1 2010. Coolaler used the Turbo Boost mode to overclock these chips. The Core i7 860S with its Turbo Boost multiplier of 26.0, reached 4.536 GHz (26 x 174.47 MHz, vCore at 1.32V), while the Core i5 750S reached 3.441 GHz (24 x 143.4 MHz, vCore at 1.32V). Both were seated on an ASUS P7P55D Deluxe motherboard, with one 2 GB DDR3 memory module.

Source: Coolaler
Add your own comment

17 Comments on Energy-Efficient Intel Core i5, Core i7 ''Lynnfield'' Processors Overclocked

#1
mdm-adph
I know it's just a test rig, but pairing that with a GeForce FX 5500 seems like some kind of joke. :wtf:

Is there some technical reason why they'd want to do that?
Posted on Reply
#2

by: mdm-adph
I know it's just a test rig, but pairing that with a GeForce FX 5500 seems like some kind of joke. :wtf:

Is there some technical reason why they'd want to do that?
its a quadro workstation card, P55 doesnt have AGP :toast:
Posted on Edit | Reply
#3
DaJMasta
I hope they don't plan on charging more for these.... you can reduce TDP by 13W just by undervolting a normal proc.... if they try to sell downclocked versions for more with such a low power savings they won't be competitive at all.


Bring on the 45W i7s!
Posted on Reply
#4
cauby
by: DaJMasta
I hope they don't plan on charging more for these.... you can reduce TDP by 13W just by undervolting a normal proc.... if they try to sell downclocked versions for more with such a low power savings they won't be competitive at all.


Bring on the 45W i7s!
Well,just compare the price of the normal q9550 an the q9550s.Looking here on newegg,the normal(95W) one sells for about U$220,while the "economic"(65W) one for U$350.Sure,it's a bigger than the 13W difference of the i5750,but still,paying 130 more for that is just out of reach.
Posted on Reply
#5
lemonadesoda
Bizarre. Just imagine how these over-turboed CPUs will behave in practice. Zooming on single threaded applications, a second thread kicks in, and the performance will throttle. Have some background tasks too, and the CPU will fall flat on its face (in relative terms).

Think of a game... throw in a little physx... and oohps... the thing just choked.

OK, so this setup will get some great superpi scores, but will be pretty poor on cinebench, geekbench, fritz, etc.
Posted on Reply
#6
Binge
Overclocking Surrealism
by: lemonadesoda
Bizarre. Just imagine how these over-turboed CPUs will behave in practice. Zooming on single threaded applications, a second thread kicks in, and the performance will throttle. Have some background tasks too, and the CPU will fall flat on its face (in relative terms).

Think of a game... throw in a little physx... and oohps... the thing just choked.

OK, so this setup will get some great superpi scores, but will be pretty poor on cinebench, geekbench, fritz, etc.
That's not how turbo works at all.
Posted on Reply
#7
btarunr
Editor & Senior Moderator
by: lemonadesoda
Bizarre. Just imagine how these over-turboed CPUs will behave in practice. Zooming on single threaded applications, a second thread kicks in, and the performance will throttle. Have some background tasks too, and the CPU will fall flat on its face (in relative terms).
If the single core decides it's not able to cope with the load, it wakes its neighbours up.
Posted on Reply
#8
WarEagleAU
Bird of Prey
I tell ya thats a nice boost on those with turbo boost and the TDP is pretty awesome.
Posted on Reply
#9
BazookaJoe
Anyone can flick a software multiplier, take a screen-shot and flip it back - By that token my Q6600 does 4.8Ghz on air cooling...

Lets see some actual BENCHMARKS proving any form of reasonable stability ?
Posted on Reply
#11
wolf
Performance Enthusiast
mmm cpu-z stable is different from gaming stable, different again from 100% stable.
Posted on Reply
#12
lemonadesoda
by: Binge
That's not how turbo works at all.
Oh, pray tell!

Coolaler was able to OC the CPU's to a very high clock WHEN IDLE! Big freaking deal.

When that CPU is doing some heavy crunching (on single thread) the turbo multiplier will cut down due to how thermal management works on the CPU. With good cooling, you might just be able to maintain that overclock. However, Coolaler didnt show such figures, and we can therefore only presume his superpi run wasnt stable, OR, the turbo cut back giving not so amazing figures.

But more, IF that CPU is running a multithread crunch, the turbo will cut back a lot more.

And that just IS how turbo mode works. Post #6. Read and repeat. For all your OC'ing effort, throw in some serious multitasking and you will be BACK AT NEAR STOCK performance.
Posted on Reply
#13
btarunr
Editor & Senior Moderator
by: lemonadesoda
Oh, pray tell!

Coolaler was able to OC the CPU's to a very high clock WHEN IDLE! Big freaking deal.
You can override Turbo. You can tell it when to go into Turbo Mode. SuperPi enthusiasts are surely going to like this, since Turbo is giving them a free multiplier boost to support higher overclocks.
Posted on Reply
#14
lemonadesoda
btarunr,

Do you know the exact mechanisms for how the [turbo+thermal managment] feature works? Is there a complex lookup table which basically manages the turbo multiplier based on core utilisation (backend and frontend instruction pipe)? Or is it fully "live and dynamic" thermally managed? Or a mix of both?

What happens when a stable idle OC (like above) gets attacked with some heavy crunch number? Will the turbo+thermal management cut back to maintain stability, and to what level, or will the CPU operate beyond the planned TDP and possibly fail? Curious.
Posted on Reply
#15
btarunr
Editor & Senior Moderator
If you've invested lot of work in getting an OC stable, you'll leave the Turbo Mode on. When the processor leaves Turbo Mode, all that's changed with respect to the settings is the multiplier value (besides of course the 3~6 more logical CPUs). So if your OC was stable at say 24 x 150 MHz (the frequency you increased), then it becomes 19 x 150 MHz.

I would imagine the processor's on-the-fly Turbo Mode management to be similar to that of the Intel SpeedStep (EIST), where the processor adjusts itself instantly depending on the load. Control over the processor's power states are given to local firmware.

Posted on Reply
#16
lemonadesoda
If it is based on a load lookup table like Speedstep (which I also presume is the case and the easiest to implement), rather than clever "true" TDP sensing circuitry, then it is possible that a "stable" turbo OC wont actually be stable on a multithread attack.

Conversely a stable multithread OC may not be stable when flying on an even high multiplier single thread.

It's a double balancing act... and it probably why we dont see coolaler showing any successful benchmarks at those speeds.

Will be interesting to see what the OC community can pull
Posted on Reply
#17
SonDa5
:rockout: This is exactly what they have to do. Now maby I'll leave my Q9550 E0.
Posted on Reply
Add your own comment